An adaptive ultrasonic wave anti-fouling method and system based on end-cloud cooperation

Abstract

The application relates to the field of marine engineering and industrial antifouling technology, and discloses a self-adaptive ultrasonic antifouling method and system based on end-cloud cooperation, a double-layer pseudo-peak elimination mechanism of dynamic window screening and slope form identification on the end side, and dynamic micro-compensation operation of a resonance point, so that dynamic and accurate tracking of a transducer resonance point is realized, triple interference is completely decoupled, the transducer always works at an optimal resonance point, the root cause of the electroacoustic conversion efficiency attenuation is avoided, the ultrasonic antifouling energy efficiency is greatly improved, temperature drift decoupling is completed based on a temperature-resistance reference surface on the cloud side, a fusion feature vector is constructed in combination with multi-dimensional load characteristics on the end side, secondary accurate determination of the load state is realized, grid abnormal data tracing and quantitative statistics are completed through sea area gridding and equipment geographical position mapping, and the problems of antifouling strategy misjudgment and regional biological outbreak early warning lag are effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of marine engineering and industrial antifouling technology, and in particular to an adaptive ultrasonic antifouling method and system based on end-to-cloud collaboration. Background Technology

[0002] Marine biofouling is a common problem faced by underwater equipment and structures such as ships, offshore platforms, and seawater cooling systems. Chemical antifouling technologies are gradually being restricted due to environmental pollution issues, while ultrasonic physical antifouling technology has become the mainstream method. Existing ultrasonic antifouling technologies primarily utilize piezoelectric transducers to emit ultrasonic waves at specific frequencies of 20kHz-40kHz. Its core mechanism involves using the cavitation effect and micro-vibrations generated during sound wave propagation in a liquid medium to disrupt the vacuolar structure of marine organisms or prevent them from secreting adhesive mucus, thereby disrupting the attachment and growth environment of marine organisms and achieving the antifouling objective. However, currently, commercial equipment using this technology generally adopts a fixed center frequency operating mode or only uses a simple open-loop frequency sweep method. It cannot adaptively adjust the ultrasonic operating parameters according to the dynamic changes in water temperature, water quality, flow velocity, marine biological community composition, and attachment status. It also lacks a real-time perception and feedback adjustment mechanism for fouling status, which easily leads to insufficient or excessive fouling prevention, resulting in large fluctuations in fouling prevention efficiency and high equipment energy consumption. At the same time, each fouling prevention device operates independently without coordinated control capabilities, making it difficult to adapt to the differentiated fouling prevention needs of different application scenarios. This restricts the practical application effect and large-scale promotion of ultrasonic fouling prevention technology.

[0003] The above-mentioned and existing related technologies often have the following drawbacks: 1. Due to the dual drift interference of thermal drift and water load drift in the dynamic marine environment, the existing frequency sweeping algorithm lacks the ability to remove pseudo-peaks through dynamic window filtering and slope morphology recognition, and cannot resist the interference of static capacitance pseudo-peaks. At the same time, the existing technology lacks a collaborative decoupling and dynamic control mechanism for this triple interference, which makes it impossible to decouple the triple interference at the same time. It is difficult to achieve dynamic and accurate tracking of the transducer resonant point, which ultimately leads to the transducer deviating from the optimal resonant point for a long time, the electroacoustic conversion efficiency continuously decays, and the ultrasonic antifouling efficiency drops significantly. 2. Due to the fact that the existing technology under the edge-cloud collaborative architecture only realizes the one-way uploading and simple monitoring of single device impedance data, it has not built an abnormal data tracing mechanism for temperature drift decoupling and multi-dimensional feature fusion, nor has it linked and integrated the individual operating characteristics of the equipment with the grid-based statistics of the sea area. In addition, it lacks the logic to distinguish the impedance anomaly characteristics of individual equipment failures and regional biological outbreaks. As a result, it is impossible to accurately distinguish between individual equipment failures and regional biological outbreaks through single device impedance anomaly data. There is also no grid-based collaborative judgment technology solution that can simultaneously realize the diagnosis of individual equipment failures and the early warning of regional biological outbreaks. This is prone to misjudgment of antifouling strategies and lag in early warning of regional biological outbreaks, and it is impossible to realize the active defense of ultrasonic antifouling and predictive maintenance of equipment. Summary of the Invention

[0004] The technical problem to be solved by this invention is that in the dynamic marine environment, the transducer faces triple interference from thermal drift, water load drift and static capacitance pseudo-peaks. Existing technologies cannot simultaneously decouple these triple interferences and achieve dynamic and accurate tracking of the resonant point. To address this, we propose an adaptive ultrasonic antifouling method and system based on end-to-cloud collaboration.

[0005] To achieve the above objectives, this application adopts the following technical solution: an adaptive ultrasonic anti-fouling method and system based on end-cloud collaboration, including an end-side adaptive resonant tracking and load control process and a cloud-side temperature drift decoupling and gridded collaborative early warning process, the specific steps of which are as follows: Step S1: After the edge-side smart terminal is powered on, it performs a full-dimensional self-test and verification. After the verification is passed, it performs a coarse frequency sweep on the piezoelectric transducer, collects the original frequency-current data, and constructs the frequency-current response curve. Step S2: The end-side intelligent terminal uses dynamic window filtering to initially remove far-field pseudo-peaks in the frequency-current response curve, and then uses slope shape recognition to accurately remove near-field static capacitance pseudo-peaks, thus locking the effective mechanical resonance point of the piezoelectric transducer. Step S3: The end-side intelligent terminal performs dynamic micro-compensation operation on the effective mechanical resonance point, and tracks the resonance point offset of the piezoelectric transducer caused by thermal drift and water load drift in real time, so as to keep the transducer always working at the optimal resonance point. Step S4: The end-side intelligent terminal collects the current time-domain and frequency-domain characteristics at the optimal resonance point, constructs a multi-dimensional load feature vector, and makes a preliminary judgment on the transducer load state based on the feature vector. Step S5: Based on the preliminary judgment result of the load status, the edge intelligent terminal adaptively switches the anti-pollution working mode and adjusts the anti-pollution strategy parameters. After completing the execution of the edge anti-pollution strategy, it standardizes and packages the running data and uploads it to the cloud-side intelligent management platform on a regular basis. Step S6: The cloud-side intelligent management platform receives the operational data uploaded from the receiving end, performs protocol parsing and validity verification, and then performs standardized classification and storage. Step S7: The cloud-side intelligent management platform completes temperature drift decoupling based on the temperature-impedance reference surface, calculates the residual impedance, and combines it with the end-side multi-dimensional load feature vector to construct a cloud-side fused feature vector to achieve secondary determination of load status. Step S8: The cloud-based intelligent management platform divides the sea area into virtual grids, maps the geographical locations of the terminal devices to the corresponding grids, counts the abnormal status of devices in each grid, and completes the source determination of abnormal data. Step S9: The cloud-based intelligent management platform generates a regional collaborative pollution prevention strategy based on the source tracing judgment results, converts the strategy into control commands, and sends them in batches to the end-side devices within the target grid; Step S10: The cloud-side intelligent management platform collects the operational data after the execution of the terminal-side strategy, completes the judgment and feedback of the anti-pollution effect, realizes the self-optimization of the model and strategy based on the feedback results, and feeds the optimization results back to the terminal-side intelligent terminal.

[0006] Preferably, in step S1, the full-dimensional self-test verification includes hardware module path verification, power parameter verification, transducer operating environment verification, and communication link verification. The coarse frequency sweep includes frequency sweep parameter initialization, test waveform transmission, current data acquisition, and frequency-current response curve construction. The end-side digital signal processing module controls the power drive module to transmit a continuous sine wave test waveform according to the preset frequency sweep range. The current detection circuit synchronously acquires the effective current value at each frequency point. After filtering, the frequency-current response curve is constructed according to the correspondence between the frequency point and the effective current value.

[0007] Preferably, in step S2, the dynamic window screening includes constructing an effective bandwidth window, extracting peak points, and screening peaks within the window. An effective bandwidth window is constructed based on the nominal center frequency of the piezoelectric transducer. All local peak points of the current are extracted from the frequency-current response curve. Only peak points whose frequency points fall within the effective bandwidth window are retained. Peak points outside the window are determined to be far-field pseudo-peaks and are removed. Slope morphology recognition includes candidate peak slope calculation, true peak determination, and optimal resonance point locking. For each candidate peak point within the window, the morphology slope of its adjacent frequency points is calculated. Peak points whose morphology slope meets the preset sharpness threshold are determined as valid mechanical resonance points. The frequency point with the largest effective current value among all valid mechanical resonance points is selected as the optimal resonance point of the piezoelectric transducer.

[0008] Preferably, in step S3, dynamic micro-compensation includes compensation parameter initialization, micro-frequency sweep triggering, resonant point update, and compensation result upload. The terminal-side intelligent terminal performs a small-range micro-frequency sweep within a preset range of the optimal resonant point according to a preset cycle, and collects the effective current value and morphological slope of the micro-frequency sweep point. If there is a frequency point with a larger effective current value that meets the sharpness threshold, the optimal resonant point is immediately updated to that frequency point, and the updated resonant point parameters are simultaneously uploaded to the cloud-side intelligent management platform.

[0009] Preferably, in step S4, the current time-domain and frequency-domain feature acquisition includes high-speed sampling of the current waveform, time-domain feature extraction, and frequency-domain feature extraction. The high-speed analog-to-digital conversion module continuously acquires the bus current waveform at the optimal resonance point at a preset sampling rate. After filtering the current waveform, the digital signal processing module extracts time-domain features such as the effective value of the current and the peak factor. The time-domain signal is converted into a frequency-domain signal through fast Fourier transform, and frequency-domain features such as total harmonic distortion are extracted. The multidimensional load feature vector is an integrated vector of the effective value of current, peak factor, and total harmonic distortion. The end-side intelligent terminal compares the feature vector with the preset load state judgment threshold to make a preliminary judgment on the clean state, initial biofilm state, hard adhesion state, and equipment failure state.

[0010] Preferably, in step S5, the anti-fouling working mode includes normal anti-fouling mode, early warning enhancement mode, strong stripping mode, and fault shutdown mode. The terminal-side intelligent terminal adjusts the anti-fouling strategy parameters such as output power, drive waveform, and working sequence according to the preliminary judgment result of the load status. The operational data includes device number, optimal resonant frequency, multidimensional load characteristic vector, load status, operating mode, device temperature, measured impedance, and online status. All data is packaged in a standardized manner according to the cloud-side protocol and uploaded periodically via the wireless communication module. When the network is interrupted, it is cached locally and automatically re-uploaded after the network is restored.

[0011] Preferably, in step S7, the temperature-impedance reference surface is a nonlinear surface model of the piezoelectric transducer constructed in clean water based on temperature, frequency, and impedance. The cloud-side intelligent management platform queries the corresponding theoretical clean impedance in the model based on the equipment temperature and optimal resonant frequency uploaded by the end side, and calculates the residual impedance by the difference between the measured impedance and the theoretical clean impedance, thereby decoupling temperature drift from bioattachment and equipment failure. The cloud-side fusion feature vector is an integrated vector of residual impedance and end-side multidimensional load feature vector. The cloud-side intelligent management platform inputs the fusion feature vector into the machine learning model to achieve secondary determination of load status, and at the same time completes the accurate distinction between individual equipment faults and bio-attachment anomalies.

[0012] Preferably, in step S8, the virtual grid division of the sea area sets the grid scale according to the distribution density of equipment in the sea area, divides the sea area into equal-scale sections according to latitude and longitude, and assigns a unique grid number. The device geolocation mapping involves binding the latitude and longitude coordinates of the end device with the corresponding grid number to establish a mapping relationship between the device and the grid; The abnormal data tracing and determination includes the statistics of the number of abnormal devices in the grid, the calculation of the abnormal ratio, and the determination of the abnormal type. Based on the ratio of abnormal devices and the distribution characteristics of faulty devices, it distinguishes between individual device faults, regional biological outbreak risks, and common faults of devices in the grid.

[0013] Preferably, in steps S1 to S10, the dynamic control of the end-side resonance point and the quantitative decision-making of the cloud-side regional collaborative strategy are realized through the end-cloud collaborative adaptive pollution prevention integrated decision-making formula. The formula expression is as follows: in, This is a comprehensive decision value for edge-cloud collaborative adaptive anti-pollution. As a weighting factor for the effectiveness of end-side resonance, For cloud-side temperature drift decoupling weighting factors, For grid anomaly collaborative weighting factor, The slope of the shape. The slope sharpness threshold. The optimal effective value of the resonant current. For the effective operating current threshold, For residual impedance, The residual impedance threshold, For total harmonic distortion, The harmonic distortion threshold. The proportion of abnormal devices in the grid. The threshold for the grid early warning ratio, This is a correction factor.

[0014] Preferably, the adaptive ultrasonic anti-fouling system based on edge-cloud collaboration is used to realize the adaptive ultrasonic anti-fouling method based on edge-cloud collaboration. It includes an edge-side intelligent terminal and a cloud-side intelligent management platform. The edge-side intelligent terminal and the cloud-side intelligent management platform realize bidirectional data interaction through a wireless communication network. The edge-side intelligent terminal includes a self-test verification module, a frequency sweep acquisition module, a pseudo-peak removal module, a resonant point locking module, a dynamic compensation module, a feature extraction module, a status determination module, a strategy execution module, and a data communication module. The cloud-based intelligent management platform includes a data access module, a temperature drift decoupling module, a secondary status determination module, a grid mapping module, an anomaly tracing module, a strategy generation module, an instruction issuance module, and a feedback optimization module.

[0015] The technical effects and advantages of this invention are as follows: This invention, by constructing a two-way closed-loop core architecture of end-to-cloud adaptive resonant tracking and cloud-side temperature drift decoupled gridded collaborative early warning, precisely solves the collaborative decoupling problem of triple interference from transducer thermal drift, water load drift, and static capacitance pseudo-peaks in dynamic marine environments. It also addresses the technical pain points of difficulty in distinguishing between individual equipment faults and regional biological outbreaks, and the susceptibility to misjudgment of antifouling strategies under end-to-cloud collaboration. Its core technical features and effects are both innovative and practical: the end-to-cloud relies on a dual-layer pseudo-peak removal mechanism of dynamic window filtering and slope morphology recognition, combined with dynamic micro-compensation operation of the resonant point, to achieve dynamic and accurate tracking of the transducer resonant point. The bottom decoupling of triple interference ensures that the transducer always operates at the optimal resonance point, fundamentally avoiding the attenuation of electroacoustic conversion efficiency and significantly improving the ultrasonic antifouling efficiency. The cloud side completes temperature drift decoupling based on the temperature-impedance reference plane, and constructs a fused feature vector by combining the multi-dimensional load characteristics of the end side, realizing secondary accurate judgment of the load status. At the same time, through the grid division of the sea area and the mapping of the equipment's geographical location, the source tracing and quantitative statistics of grid anomaly data are completed, and a precise differentiation logic between individual equipment failures and regional biological outbreaks is established, forming a grid-based collaborative early warning and strategy distribution mechanism, effectively avoiding the problems of misjudgment of antifouling strategies and delayed early warning of regional biological outbreaks. This invention incorporates an edge-cloud collaborative adaptive antifouling comprehensive decision-making formula, realizing quantitative decision-making for edge-side resonant point control and cloud-side regional collaborative antifouling strategies. Combined with the antifouling effect feedback mechanism after strategy execution, it completes the self-learning and self-optimization of machine learning models and antifouling strategies, and feeds the optimization results back to the edge, forming a two-way closed-loop self-evolutionary system between the edge and cloud. At the same time, the edge-side adaptive control and cloud-side gridded collaborative technology design of this invention can be adapted to the environmental characteristics of different marine application scenarios such as offshore wind power bases, aquaculture cages, and ocean-going vessels, realizing active defense and predictive maintenance of ultrasonic antifouling, taking into account both precise control of single equipment and regional collaborative antifouling, significantly improving the intelligence, collaboration, and precision of ultrasonic physical antifouling of marine engineering equipment, and without the use of chemical agents throughout the process, meeting the requirements of green and environmentally friendly marine ecological protection, and has significant engineering application value and ecological benefits. Attached Figure Description

[0016] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts: Figure 1 This is a schematic diagram of the process of the adaptive ultrasonic antifouling method based on edge-cloud collaboration of the present invention; Figure 2 This is a schematic diagram of the end-side hardware device of the adaptive ultrasonic anti-fouling system based on end-cloud collaboration of the present invention. Detailed Implementation

[0017] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0018] Example 1 Reference Figure 1 , Figure 2 As shown, this invention provides a technical solution: an adaptive ultrasonic anti-fouling method based on end-to-cloud collaboration, including an end-side adaptive resonant tracking and load control process and a cloud-side temperature drift decoupling and gridded collaborative early warning process. The end-side adaptive resonant tracking and load control process provides the cloud-side temperature drift decoupling and gridded collaborative early warning process with full-dimensional end-side operational data, while the cloud-side temperature drift decoupling and gridded collaborative early warning process provides the end-side adaptive resonant tracking and load control process with precise decision-making instructions and optimization parameters. The two form a two-way closed-loop linkage between the end and cloud. The specific steps are as follows: Step S1: After the edge-side smart terminal is powered on, it performs a full-dimensional self-test and verification. After the verification is passed, it performs a coarse frequency sweep on the piezoelectric transducer, collects the original frequency-current data, and constructs the frequency-current response curve.

[0019] After the smart terminal on the edge is powered on, the self-test and verification module immediately starts a full-dimensional self-test and verification. The verification content includes hardware module path verification, power parameter verification, piezoelectric transducer working environment verification, and communication link verification. The hardware module path verification performs path testing on core modules such as digital signal processing module, high-speed analog-to-digital conversion module, power drive module, and current detection module to ensure that the modules work normally. Power parameter verification checks whether the voltage and frequency of the input power supply meet the rated operating requirements of the equipment to avoid equipment failure caused by overvoltage or undervoltage; piezoelectric transducer working environment verification determines whether the transducer is in a normal water working environment through current detection to avoid damage to the transducer caused by dry running without water. The communication link verification detects the connection status between the wireless communication module and the cloud-based intelligent management platform, and verifies the validity of the device number.

[0020] After all verification items pass, the self-test verification module sends a start command to the frequency sweep acquisition module. If any verification item fails, the corresponding protection mechanism is immediately triggered and a fault warning is sent to the cloud side, and the system is shut down for manual troubleshooting.

[0021] After receiving the start command, the frequency sweep acquisition module initializes the frequency sweep parameters, sets the frequency sweep range and sweep step of the piezoelectric transducer, and controls the power drive module to transmit a continuous sine wave test waveform according to the initialization parameters. Each frequency point is continuously transmitted for a preset duration to ensure the stability of current data acquisition. The current detection circuit synchronously acquires the effective value of the bus current at each frequency point, filters the acquired current data to remove high-frequency noise interference, and then constructs a frequency-current response curve according to the one-to-one correspondence between the frequency point and the effective value of the current. The curve data is stored in the local buffer on the end side to provide raw data for subsequent spurious peak removal.

[0022] Step S2: The end-side intelligent terminal uses dynamic window filtering to initially remove far-field spurious peaks in the frequency-current response curve, and then uses slope morphology recognition to accurately remove near-field static capacitance spurious peaks, thus locking the effective mechanical resonance point of the piezoelectric transducer.

[0023] The pseudo-peak removal module retrieves the frequency-current response curve from the local cache and first performs a dynamic window filtering operation, using the nominal center frequency of the piezoelectric transducer. Based on this, construct an effective bandwidth window. The algorithm extracts all local peak points of current from the frequency-current response curve, i.e., points where the effective current value at a certain frequency is greater than the effective current values ​​of its left and right adjacent frequencies, forming a set of peak points. This set is then iterated through, retaining only peak points whose frequencies fall within the effective bandwidth window. Peak points outside the window are identified as far-field spurious peaks caused by cable inductance or external electromagnetic interference and are removed to reduce the computational load of subsequent algorithms. If no peak points are found within the effective bandwidth window, the window is expanded by a preset range and re-filtered. If no peak points are still found, it is determined that the transducer load is abnormal, an early warning is sent to the cloud side, and the frequency sweep acquisition module is triggered to re-sweep the frequency.

[0024] After completing the dynamic window filtering, the pseudo-peak removal module performs slope morphology recognition, identifying each candidate peak point within the window. Calculate its left and right adjacent frequency points ( ) morphological slope The slope of the current peak reflects the rising trend of the current peak and is a core parameter for distinguishing between the effective mechanical resonant point and the static capacitance pseudo-peak. The calculation formula is as follows: in, The effective value of the current at the candidate peak point. The effective value of the current at the frequency points adjacent to the candidate peak point. This is a frequency sweep step.

[0025] The calculated shape slope Compared with the preset sharpness threshold In comparison, if The point is determined to be an effective mechanical resonance point, and the corresponding current change is a sharp rise caused by the mechanical resonance of the piezoelectric transducer. like Near-field spurious peaks caused by static capacitance effects are identified and removed. The resonant point locking module selects the frequency point with the largest effective current value from all valid mechanical resonant points as the optimal resonant point for the piezoelectric transducer. It sends a command to the power drive module to lock the frequency point as the normal operating frequency of the end side.

[0026] Step S3: The end-side intelligent terminal performs dynamic micro-compensation operation on the effective mechanical resonance point, and tracks the resonance point offset of the piezoelectric transducer caused by thermal drift and water load drift in real time, so as to keep the transducer always working at the optimal resonance point.

[0027] After receiving the optimal resonant point parameters from the resonant point locking module, the dynamic compensation module initializes the compensation parameters and sets the period of dynamic micro-compensation. Scope of Compensation With compensation step The micro-frequency sweep operation is triggered according to the preset compensation cycle. During each micro-frequency sweep, the transmission of the normal anti-pollution waveform is paused, and the transmission is performed within the preset compensation range of the current optimal resonance point. Perform a small-range micro-sweep, collect the effective current value and morphological slope of the micro-sweep frequency point, and if a certain micro-sweep frequency point exists... satisfy and Immediately update the optimal resonant point to this frequency, completing the dynamic compensation of the resonant point, that is... ; If the frequency does not exist, keep the current optimal resonant point unchanged.

[0028] After compensation is completed, the dynamic compensation module uploads the updated optimal resonant point parameters to the cloud-based intelligent management platform in real time, and at the same time updates the local operating parameters on the end side to ensure that the piezoelectric transducer always works at the optimal resonant point, decoupling the resonant point offset interference caused by thermal drift and water load drift, and maintaining the maximum electroacoustic conversion efficiency.

[0029] Step S4: The end-side intelligent terminal collects the current time-domain and frequency-domain characteristics at the optimal resonance point, constructs a multi-dimensional load feature vector, and makes a preliminary judgment on the transducer load state based on the feature vector.

[0030] When the piezoelectric transducer is operating at its optimal resonant point, the feature extraction module initiates high-speed sampling of the current waveform, and the high-speed analog-to-digital conversion module operates at a preset sampling rate. The system continuously acquires bus current waveforms, storing the data for a preset duration each time. The digital signal processing module filters the acquired current waveforms and extracts the RMS current value. Peak factor The isochronous domain characteristics, where the peak factor is the ratio of the peak current to the effective current, are used to quantify the intensity of the ultrasonic cavitation effect. The calculation formula is as follows: in, This represents the peak value of the current waveform.

[0031] Subsequently, a Fast Fourier Transform is used to convert the time-domain signal of the current into a frequency-domain signal, and the total harmonic distortion is extracted. The equal-frequency domain characteristics and total harmonic distortion reflect the nonlinearity of bioburden and are used to quantify the attachment degree of marine organisms. The calculation formula is as follows: in, This is the effective value of the fundamental current. for Effective value of the second harmonic current.

[0032] The feature extraction module integrates the RMS current value, peak factor, and total harmonic distortion to construct a multi-dimensional load feature vector at the end side. The feature vectors are stored in the local cache to provide a quantitative basis for load status determination.

[0033] The status determination module retrieves the multi-dimensional load feature vector from the local cache and the preset load status determination threshold, and uses the anti-pollution risk preliminary judgment index. The preliminary determination of the load state of the piezoelectric transducer is achieved using the following calculation formula: in, The weighting coefficients and , The threshold range of the effective value of the current. The peak factor under clean conditions. The peak factor is under hard adhesion conditions. This represents the maximum threshold for total harmonic distortion.

[0034] The load status is divided into clean state, initial biofilm state, hard adhesion state, and equipment failure state. Among them, the equipment failure state is judged first. Regardless of other characteristic parameters, it is directly determined to be a fault state of the equipment, covering fault types such as transducer detachment, loose bonding, and dry burning; according to The range of values ​​matches the corresponding biological attachment state. The higher the value, the greater the degree of biological adhesion.

[0035] The status determination module stores the determination results in the local cache and sends the status determination results to the cloud side. If the determination is in the initial biofilm state, hard attachment state, or equipment failure state, it immediately sends an abnormal warning to the cloud side, along with a multi-dimensional load feature vector and the determination result.

[0036] Step S5: Based on the preliminary judgment result of the load status, the edge intelligent terminal adaptively switches the anti-pollution working mode and adjusts the anti-pollution strategy parameters. After completing the execution of the edge anti-pollution strategy, it packages the running data in a standardized manner and uploads it to the cloud-side intelligent management platform on a regular basis.

[0037] The strategy execution module retrieves the preliminary load status assessment results from the status assessment module and matches the corresponding anti-pollution operating mode based on the assessment results. The anti-pollution operating modes include normal anti-pollution mode, early warning enhancement mode, strong stripping mode, and fault shutdown mode. Different operating modes correspond to different output power. Drive waveform, operating timing parameter.

[0038] The normal anti-fouling mode is adapted to clean conditions, and adopts low power, continuous sine wave, and normal working sequence to achieve low power consumption anti-fouling. The early warning enhancement mode adapts to the initial state of biofilm, increases output power, densifies the working sequence, maintains a continuous sine wave, and inhibits the formation and development of biofilm; The powerful peeling mode is adapted to hard attachment conditions. It adopts full power, variable frequency pulse train and encrypted working timing, and uses strong cavitation effect to destroy the attachment interface of hard marine organisms to achieve peeling and antifouling. The fault shutdown mode adapts to equipment fault conditions, immediately cutting off power output, stopping ultrasonic emission, and triggering shutdown protection.

[0039] The strategy execution module sends instructions to the power drive module according to the matched working mode to adjust the output power, drive waveform and working timing, so as to realize the adaptive switching of the anti-pollution strategy. At the same time, it monitors the device operating parameters after the strategy is executed in real time. If the parameters exceed the hardware rated range, they are immediately reduced to the threshold and an early warning is sent to the cloud side.

[0040] The data communication module retrieves all operational data from the terminal side, including the device number and the optimal resonant frequency. Multidimensional load feature vector Load status, operating mode, output power Equipment temperature Measured impedance The online status and other parameters are standardized and packaged according to the communication protocol of the cloud-side intelligent management platform to generate terminal-side data packets. The data communication module is set to upload the terminal-side data packets to the cloud side through the wireless communication module. If the wireless communication network is interrupted, the data packets are stored in the local cache on the terminal side. The cache capacity meets the storage requirements for a preset duration. After the network is restored, the cached data packets are automatically re-uploaded in timestamp order to ensure the integrity of cloud-side data. At the same time, the module receives control commands and optimization parameters issued by the cloud side and synchronizes them with each module on the terminal side.

[0041] Step S6: The cloud-based intelligent management platform receives the operational data uploaded from the receiving end, performs protocol parsing and validity verification, and then performs standardized classification and storage.

[0042] The cloud-side data access module receives end-side data packets uploaded by various end-side smart terminals through an IoT access gateway, parses the data packets according to a preset communication protocol, and extracts the device number and optimal resonant frequency. Multidimensional load feature vector Equipment temperature Measured impedance Core operational data, etc.

[0043] The parsed runtime data is then validated to remove invalid data such as missing parameters, values ​​outside the reasonable range, or abnormal timestamps. A data reception confirmation or failure signal is then sent to the endpoint, with the failure signal accompanied by a specific reason for the failure.

[0044] The data access module categorizes valid data by device number and stores it in the cloud-side time-series database using timestamps as indexes. This achieves standardized data classification and storage, supports rapid data retrieval and multi-dimensional searching, and updates the online status of the cloud platform device list in real time based on the data packet reception. It also marks the last data upload time for offline devices, providing a complete and standardized data source for subsequent cloud-side algorithm implementation.

[0045] Step S7: The cloud-side intelligent management platform completes temperature drift decoupling based on the temperature-impedance reference plane, calculates the residual impedance, and combines it with the multi-dimensional load feature vector on the end side to construct a cloud-side fused feature vector to realize secondary determination of load status.

[0046] The temperature drift decoupling module retrieves the device temperature uploaded from the end side from the time-series database. Optimal resonant frequency Compared with the measured impedance Simultaneously, a preset temperature-impedance reference surface model is retrieved. This model is a nonlinear surface model of the piezoelectric transducer in clean water, constructed based on three dimensions: temperature, frequency, and impedance. It accurately reflects the impedance changes caused by the thermal drift of the piezoelectric transducer and variations in water temperature. The temperature drift decoupling module, based on the equipment temperature and the optimal resonant frequency, queries the theoretical clean impedance under the corresponding operating conditions in the temperature-impedance reference surface model. The residual impedance is calculated by the difference between the measured impedance and the theoretical clean impedance. This achieves complete decoupling between temperature drift and biofouling / equipment malfunction. The calculation formula is as follows: like Then it is corrected to The residual impedance is stored in the time series database to provide core parameters for subsequent secondary state determination.

[0047] The secondary state determination module retrieves the residual impedance from the timing database. With end-side multidimensional load feature vector By integrating the two, a cloud-side fused feature vector is constructed. The fused feature vector combines the impedance characteristics after temperature drift decoupling with the current characteristics on the end side, improving the accuracy of state determination.

[0048] The secondary state determination module integrates feature vectors input into a pre-trained machine learning model, and uses a secondary pollution risk determination index. To achieve secondary determination of load status, the calculation formula is as follows: in, The weighting coefficients and , The maximum threshold value for residual impedance. This is a preliminary risk assessment index for end-side contamination.

[0049] The secondary assessment results include health status, initial biofilm status, hard biofilm attachment status, and individual equipment malfunctions, while accurately distinguishing between individual equipment malfunctions and abnormal biofilm attachment. The secondary assessment module compares the secondary assessment results with the initial assessment results from the end-side; if the two match, the assessment is considered accurate. If the two are inconsistent, the cloud-side secondary judgment result shall prevail. The judgment result shall be stored in the time series database, and the judgment result and policy adjustment instruction shall be sent to the corresponding edge-side smart terminal. The edge side shall immediately adjust the anti-pollution policy after receiving the instruction.

[0050] Step S8: The cloud-based intelligent management platform divides the sea area into virtual grids, maps the geographical locations of the end devices to the corresponding grids, counts the abnormal status of devices in each grid, and completes the source determination of abnormal data.

[0051] The grid mapping module sets the scale of the virtual grid in the sea area based on the distribution density of end-side devices within the sea area. It divides the pollution prevention sea area into equal-scale subdivisions according to latitude and longitude, assigning a unique grid number to each grid to form a virtual grid map of the sea area. Subsequently, it retrieves the latitude and longitude coordinates of all end-side smart terminals from the cloud-side device library, binds the device number of each end-side device to its corresponding grid number, establishes a mapping relationship between device number, grid number, and latitude and longitude coordinates, generates a device list for each grid, and counts the total number of devices within each grid. With the number of online devices The mapping relationship and grid device list are stored in the database, and a visualized gridded electronic map of the sea area is generated on the cloud platform interface to display the device distribution and online status of each grid in real time, providing a spatial carrier for gridded collaborative early warning.

[0052] The anomaly tracing module retrieves the secondary load status determination results of all end-side devices in each grid from the time-series database, performs statistics by grid number, and counts the number of abnormal devices in each grid. Number of faulty devices Calculate the proportion of abnormal equipment The calculation formula is: Based on the distribution characteristics of faulty equipment, the source of abnormal data is determined, and three types of abnormalities are distinguished: like and Scattered distribution indicates individual equipment failure; like and Scattered distribution, judged as a regional biological outbreak risk, according to The value is further divided into initial warning ( ) and outbreak warning ( ); like If the faults are concentrated within a single grid, they are identified as common faults among devices within that grid.

[0053] in, The threshold for the grid early warning ratio, The threshold values ​​for the secondary assessment index of pollution risk are used. The anomaly tracing module marks the tracing results on a gridded electronic map of the sea area, with different map marking colors corresponding to different anomaly types, thus achieving a visual display of the anomaly status.

[0054] Step S9: The cloud-based intelligent management platform generates a regional collaborative pollution prevention strategy based on the source tracing judgment results, converts the strategy into control commands, and sends them in batches to the end-side devices within the target grid.

[0055] The strategy generation module retrieves the grid anomaly type and source determination results from the anomaly tracing module. Based on the preset strategy generation rules, it matches corresponding regional collaborative anti-fouling strategies for different anomaly types. The strategy generation rules are personalized by combining the grid's biofouling level and the proportion of abnormal equipment. The higher the proportion of abnormal equipment and the higher the biofouling level, the stronger the anti-fouling strategy.

[0056] For early warnings of regional biological outbreaks, a coordinated early warning strategy is implemented to enhance the anti-fouling power of equipment within the grid and increase the working sequence. For early warnings of regional biological outbreaks, a strong stripping coordinated strategy is implemented to activate the full-power, variable-frequency pulse train working mode of equipment within the grid. For common faults of equipment within the grid, a fault emergency strategy is implemented to trigger the shutdown protection of equipment within the grid and send batch fault warnings to maintenance personnel.

[0057] The strategy generation module transforms the matched regional collaborative anti-fouling strategy into standardized control instructions that can be executed by the edge intelligent terminal. The control instructions include core parameters such as working mode, output power, drive waveform, and working sequence. According to the anomaly type, they are packaged into grid batch instruction packages by grid number or into single device precise instruction packages by device number. Regional biological outbreak risks and common faults of devices within the grid are handled by grid batch instruction packages, while individual device faults are handled by single device precise instruction packages.

[0058] The command delivery module sends control command packets to edge smart terminals within the target mesh via a wireless communication network. Batch delivery supports concurrent delivery to multiple devices, improving command delivery efficiency. For edge devices that are offline, the control command packets are cached in the cloud-based command library and resent immediately upon device re-entry. Simultaneously, the module receives command execution feedback signals from the edge devices and calculates the command execution success rate. The calculation formula is: in, This represents the number of devices that successfully executed the command. The target number of devices. If... (Preset success rate threshold) Send command execution abnormality warning to maintenance personnel.

[0059] Step S10: The cloud-side intelligent management platform collects the operational data after the execution of the terminal-side strategy, completes the judgment and feedback of the anti-pollution effect, realizes the self-optimization of the model and strategy based on the feedback results, and feeds the optimization results back to the terminal-side intelligent terminal.

[0060] The feedback optimization module collects operational data from the time-series database after the target grid's end-side devices have implemented the collaborative anti-pollution strategy for a preset duration, and extracts the residual impedance. Total Harmonic Distortion Secondary assessment index of pollution prevention risk By comparing the changes in key performance indicators before and after the strategy implementation, the improvement rate of antifouling effectiveness can be calculated. The anti-fouling effect is determined by the following formula: in, Core performance indicators ( ), This refers to the number of indicators.

[0061] Judgment rule: If (Preset improvement rate threshold) If the anti-fouling is deemed effective, the current anti-fouling strategy will continue to be implemented; like If the anti-fouling effect is deemed inadequate; It was determined to be a failure of anti-fouling measures.

[0062] In response to situations where anti-fouling effects are poor or fail, the power, timing, and other parameters of the anti-fouling strategy are adjusted to generate an optimized anti-fouling strategy, which is then redistributed to the edge by the strategy generation module and the command distribution module.

[0063] The feedback optimization module takes the edge-side runtime data after strategy execution as incremental samples and inputs them into the cloud-side machine learning model to retrain and optimize the model, adjusting the model's weight coefficients. By adjusting the judgment threshold, the model's adaptability to different sea areas and different biological attachment characteristics is improved, enabling the model to self-learn and self-optimize. Simultaneously, based on the judgment results of antifouling effectiveness, the cloud-based antifouling strategy generation rules are optimized and iterated, the strategy parameter library is updated, and the accuracy and effectiveness of collaborative antifouling strategies are improved.

[0064] The feedback optimization module feeds back the optimized machine learning model's judgment threshold and anti-pollution strategy parameters to the corresponding edge smart terminal through the data communication module, updating the judgment threshold and strategy parameters on the edge side, and realizing self-optimization through a two-way closed loop between the edge and the cloud.

[0065] Furthermore, throughout the entire operation process from steps S1 to S10 above, the edge-cloud collaborative adaptive pollution prevention integrated decision-making formula is used. This enables dynamic control of the resonant point of the end-side piezoelectric transducer and quantitative decision-making for collaborative anti-pollution strategies in the cloud-side region. The formula expression is as follows: in, This is a comprehensive decision value for edge-cloud collaborative adaptive anti-pollution. As a weighting factor for the effectiveness of end-side resonance, For cloud-side temperature drift decoupling weighting factors, For grid anomaly collaborative weighting factor, The slope of the shape. The slope sharpness threshold. The optimal effective value of the resonant current. For the effective operating current threshold, For residual impedance, The residual impedance threshold, For total harmonic distortion, The harmonic distortion threshold. The proportion of abnormal devices in the grid. The threshold for the grid early warning ratio, This is a correction factor.

[0066] according to The range of values ​​outputs the quantitative decision results: like End-side locking of the current The cloud side determines that the problem is caused by a single device, and then executes a precise anti-pollution strategy for that single device. like End-side fine-tuning The cloud-based system was identified as an early warning of a regional biological outbreak, and a coordinated strategy to strengthen early warning was implemented. like : Frequency of switching to strong stripping mode on the end side The cloud-based system determined that a regional biological outbreak was imminent and issued an early warning, triggering a powerful stripping and synergistic strategy. like The device is shut down on the end side. If the cloud side determines that the device is faulty, a fault warning will be sent.

[0067] Example 2 The edge-cloud collaborative adaptive ultrasonic antifouling system is used to implement the edge-cloud collaborative adaptive ultrasonic antifouling method of Example 1. It includes an edge-side intelligent terminal and a cloud-side intelligent management platform. The edge-side intelligent terminal and the cloud-side intelligent management platform realize bidirectional data interaction through a wireless communication network. The edge-side intelligent terminal is deployed at the antifouling points of marine engineering equipment such as ships, offshore wind power bases, aquaculture cages, and submarine pipelines. The cloud-side intelligent management platform is deployed on a cloud server and supports concurrent access and global management of multiple edge devices.

[0068] The edge-side intelligent terminal includes a self-testing and verification module, a frequency sweep acquisition module, a pseudo-peak removal module, a resonant point locking module, a dynamic compensation module, a feature extraction module, a status determination module, a strategy execution module, and a data communication module. These modules are integrated within the ultrasonic anti-fouling controller and work in conjunction with hardware units such as piezoelectric transducers, current detection sensors, temperature sensors, and wireless communication modules. The self-test verification module is used to perform full-dimensional self-test verification of the edge-side smart terminal, including verification of hardware module paths, power parameters, transducer working environment, and communication links, triggering protection mechanisms and sending fault warnings. The frequency sweep acquisition module is used to perform coarse frequency sweeping on the piezoelectric transducer, acquire raw frequency-current data, and construct frequency-current response curves; The pseudo-peak removal module is used to achieve dual-layer removal of far-field pseudo-peaks and near-field static capacitance pseudo-peaks through dynamic window filtering and slope morphology recognition. The resonant point locking module is used to calculate the morphological slope of candidate peak points and lock the optimal resonant point of the piezoelectric transducer after comparing it with the sharpness threshold. The dynamic compensation module is used to perform dynamic micro-compensation for the optimal resonance point. It triggers micro-frequency sweep according to a preset cycle, tracks and updates the resonance point in real time, and decouples thermal drift and water load drift interference. The feature extraction module is used to acquire the current waveform at the optimal resonance point at high speed, extract time-domain and frequency-domain features, and construct a multi-dimensional load feature vector at the end side. The status determination module is used to calculate the initial judgment index of antifouling risk based on the multi-dimensional load feature vector, and compare it with the preset threshold to realize the initial judgment of the load status, distinguishing between clean, biofouling, equipment failure and other states. The strategy execution module is used to adaptively switch the anti-pollution working mode and adjust strategy parameters such as output power, drive waveform, and working timing based on the load status determination result. The data communication module is used to standardize and package the terminal-side running data and upload it to the cloud side, and to receive control commands and optimization parameters issued by the cloud side to realize terminal-cloud data interaction.

[0069] The cloud-based intelligent management platform includes a data access module, a temperature drift decoupling module, a secondary status determination module, a grid mapping module, an anomaly tracing module, a strategy generation module, a command issuance module, and a feedback optimization module. Each module is built on a cloud server and an IoT platform, supporting visual operation and remote management. The data access module is used to receive data packets uploaded by the receiving end, complete protocol parsing, validity verification and standardized classification and storage, and update the online status of the device; The temperature drift decoupling module is used to query the theoretical clean impedance and calculate the residual impedance based on the temperature-impedance reference surface model, thereby achieving decoupling of temperature drift. The secondary status determination module is used to construct cloud-side fusion feature vectors, calculate the secondary determination index of anti-fouling risk, input the data into the machine learning model to realize secondary determination of load status, and accurately distinguish between biofouling and individual equipment failures. The grid mapping module is used to complete the virtual grid division of the sea area, establish the geographical location mapping relationship between the end-side equipment and the grid, and generate a visual grid map; The anomaly tracing module is used to statistically analyze the abnormal status of equipment in each grid, calculate the proportion of abnormal equipment, complete the anomaly data tracing and determination, and distinguish between individual equipment failures and regional biological outbreaks. The strategy generation module is used to generate personalized regional collaborative pollution prevention strategies based on the source tracing results, and transform them into standardized control instructions; The instruction delivery module is used to send control instructions in batches or precisely to the end devices, cache offline device instructions, and count the instruction execution success rate. The feedback optimization module is used to collect data on the effect of strategy execution, calculate the improvement rate of anti-fouling effect to complete the anti-fouling effect judgment, realize the self-optimization of machine learning model and anti-fouling strategy based on the effect data, and feed the optimization results back to the edge.

[0070] The system achieves decoupling of triple interference of piezoelectric transducers and dynamic and accurate tracking of resonant points through the coordinated work of various modules on the end side, ensuring the energy efficiency of ultrasonic anti-fouling; and achieves abnormal data tracing and regional grid-based collaborative early warning through the coordinated work of various modules on the cloud side, avoiding misjudgment of anti-fouling strategies and delayed early warning. Through bidirectional data interaction and closed-loop linkage between the edge and cloud modules, the entire system can achieve self-learning, self-optimization, and self-evolution, adapting to dynamic and complex marine environments and realizing proactive defense against ultrasonic fouling and predictive maintenance of equipment.

[0071] Example 3 Offshore wind turbine bases are fixed-point equipment at sea, which are significantly affected by tides and diurnal variations in water temperature. Piezoelectric transducers are prone to thermal drift and water load drift. Furthermore, marine organisms attached to the bases exhibit regional aggregation characteristics, which can easily lead to localized biological outbreaks. This makes equipment operation and maintenance difficult and necessitates the implementation of unattended intelligent anti-fouling systems.

[0072] Implementation process: After the edge-side intelligent terminal is powered on, it completes self-test verification, performs a coarse frequency sweep of the transducer, and filters through a dynamic window. Remove far-field spurious peaks and calculate the shape slope ( Locking in the optimal resonance point ; Due to tidal fluctuations causing the water temperature to rise from 18℃ to 25℃ and changes in water flow velocity, the dynamic compensation module performs a micro-frequency sweep every 10 minutes. Detected at 20.8 kHz within the range. ,Will Updated to 20.8kHz to achieve dynamic tracking of the resonant point; The feature extraction module collected ,calculate It was initially determined to be in the initial biofilm state, and the end-side was switched to the early warning enhancement mode; After receiving data from the cloud-side receiver, the temperature-impedance reference surface model is queried to obtain... , combined calculate ; Build ,calculate The second determination indicates that it is in the initial biofilm state; The grid mapping module assigned the wind turbine base to a grid, and statistics showed that 7 out of 20 devices within the grid were in the initial biofilm stage. This was determined to be an early warning sign of a regional biological outbreak. The cloud-based early warning system enhances collaborative strategies, and is batch-distributed to 20 devices in the grid, with a command execution success rate of 95%. After 24 hours of execution, the cloud-based data collection results are obtained. An average decrease of 40% An average decrease of 35% was deemed effective in preventing contamination. This data was used as incremental samples to optimize the model and update the weight coefficients. And feed back to the end side.

[0073] It achieves dynamic and precise tracking of the resonant point of the wind turbine base transducer, maintains an electroacoustic conversion efficiency of over 90%, effectively suppresses biofouling within the grid, prevents regional biofouling outbreaks, and realizes unattended intelligent pollution prevention.

[0074] Example 4 Port aquaculture cages are densely located equipment. The water around the cages is eutrophic, and marine organisms (barnacles, mussels) attach quickly, which can easily lead to regional biological outbreaks. Transducers are prone to abnormal loads due to the large amount of organisms attaching. In addition, there is a lot of electromagnetic interference in the aquaculture area, and frequency sweeps are prone to false peaks.

[0075] Implementation process: When the edge-side intelligent terminal performs a coarse frequency scan, it is affected by electromagnetic interference from the water pump in the aquaculture area, resulting in a far-field spurious peak at 35kHz. This peak is filtered through a dynamic window. ( Successfully removed, locked at 20.5kHz. ( ); Barnacles quickly attach around the aquaculture cages, and samples are collected from the ends. , It was initially determined to be in a hard adhesion state, and the end side switched to a strong peeling mode; Cloud-side calculation , The second determination indicated a hard-attached state. Of the 30 devices within the grid of this aquaculture area, 12 were in a hard-attached state. This was determined to be a regional biological outbreak warning; A powerful stripping and collaboration strategy was generated on the cloud side and distributed in batches to 30 devices in the grid, with a command execution success rate of 98%. After 48 hours of execution, the cloud-based data collection results were obtained. An average decrease of 55% An average decrease of 50% indicates effective anti-fouling measures; the strategy parameters are then optimized and fed back to the end-user. One transducer within the grid was detected If the problem is directly identified as a device malfunction, the cloud side issues a precise instruction to the device, and maintenance personnel replace it in a timely manner to avoid affecting the anti-fouling of surrounding cages.

[0076] The removal rate of hard marine organisms attached to the grid reached over 85%, successfully curbing regional biological outbreaks. The equipment fault identification accuracy rate was 100%, the antifouling effect of the aquaculture cages was significant, and the survival rate of aquatic products increased by 15%.

[0077] Example 5 Ocean-going vessels are mobile point devices that cross different sea areas (East China Sea, South China Sea, Indian Ocean) during navigation. The water temperature, salinity and current velocity vary greatly, the transducer load drift is significant, and the electromagnetic interference in the ship's engine room is complex. It is necessary to achieve dynamic tracking of resonant points and adaptive anti-fouling across sea areas.

[0078] Implementation process: When a vessel is navigating in the East China Sea, end-side locking is required. After sailing to the South China Sea, the water temperature rose to 28℃ and the salinity to 33‰. The dynamic compensation module then performed a micro-frequency sweep. Updated to 20.8kHz, while maintaining an electroacoustic conversion efficiency of over 90%; When the ship was transiting a certain area of ​​the Indian Ocean, there was a large amount of plankton in the surrounding area, and samples were collected from the end side. , It was initially determined to be in the initial biofilm state, and the end-side was switched to the early warning enhancement mode; The cloud-based system updates the grid mapping in real time based on the ship's latitude and longitude, assigns the ship to a temporary moving grid, and collects statistics on abnormal statuses of ship equipment in the surrounding sea area. The system was determined to be a single device malfunction, and a precise anti-contamination strategy was implemented. If a network interruption occurs during ship navigation, the terminal will cache the operating data locally for 72 hours. After the network is restored, the data will be automatically retransmitted. After receiving the data, the cloud side will complete the status determination and policy optimization. There is no interruption due to pollution prevention. During the entire voyage, the end-side dynamic compensation module was updated a total of 12 times, always maintaining the optimal resonance point, the antifouling strategy adaptively switches according to the marine organism attachment status, with no obvious biological attachment.

[0079] It enables dynamic tracking of transducer resonant points across ocean-going vessels, adapts to changes in the aquatic environment of different sea areas, maintains electroacoustic conversion efficiency above 88%, and ensures no obvious marine organisms adhere to the ship's hull, thus achieving adaptive intelligent antifouling for mobile devices.

[0080] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. An adaptive ultrasonic antifouling method based on edge-cloud collaboration, characterized in that, The process includes end-side adaptive resonant tracking and load control, and cloud-side temperature drift decoupling and mesh-based collaborative early warning. The specific steps are as follows: Step S1: After the end-side smart terminal completes the full-dimensional self-test and verification, it performs a coarse frequency sweep on the piezoelectric transducer, collects the original frequency-current data, and constructs the frequency-current response curve. Step S2: At the end side, the far-field spurious peaks are filtered out through a dynamic window, and the near-field static capacitance spurious peaks are then identified through slope morphology to lock the effective mechanical resonance point of the piezoelectric transducer. Step S3: Perform dynamic micro-compensation on the effective mechanical resonant point at the end side, track the resonant point offset caused by thermal drift and water load drift in real time, and maintain the transducer working at the optimal resonant point. Step S4: Collect the time-domain and frequency-domain characteristics of the current at the optimal resonance point at the end side, construct a multi-dimensional load feature vector, and complete the preliminary determination of the transducer load status; Step S5: The terminal side adaptively switches the anti-pollution working mode and adjusts the strategy parameters based on the preliminary judgment result of the load status, and uploads the standardized packaged operation data to the cloud-side intelligent management platform on a regular basis. Step S6: The runtime data uploaded by the cloud-side receiving end is parsed and verified, and then standardized, classified, and stored. Step S7: The cloud side completes temperature drift decoupling based on the temperature-impedance reference surface, calculates the residual impedance, and constructs a fused feature vector by combining the multi-dimensional features of the end side to complete the secondary determination of the load status. Step S8: The cloud side performs virtual grid division of the sea area, completes the geographical location mapping of the end-side equipment, counts the abnormal status of the equipment within the grid, and completes the source determination of abnormal data. Step S9: The cloud side generates a regional collaborative anti-pollution strategy based on the source tracing judgment results, converts it into control commands, and sends them in batches to the end-side devices in the target grid; Step S10: The cloud side collects the running data after the execution of the end-side strategy, completes the judgment and feedback of the anti-pollution effect, realizes the self-optimization of the model and strategy based on the feedback results, and feeds the optimization results back to the end side.

2. The adaptive ultrasonic anti-fouling method based on edge-cloud collaboration according to claim 1, characterized in that: In step S1, the full-dimensional self-test verification includes hardware module path verification, power parameter verification, transducer operating environment verification, and communication link verification. The coarse frequency sweep includes frequency sweep parameter initialization, test waveform transmission, current data acquisition, and frequency-current response curve construction. The end-side digital signal processing module controls the power drive module to transmit a continuous sine wave test waveform according to the preset frequency sweep range. The current detection circuit synchronously acquires the effective current value at each frequency point. After filtering, the frequency-current response curve is constructed according to the correspondence between the frequency point and the effective current value.

3. The adaptive ultrasonic antifouling method based on edge-cloud collaboration according to claim 1, characterized in that: In step S2, the dynamic window filtering includes constructing an effective bandwidth window, extracting peak points, and filtering peaks within the window. An effective bandwidth window is constructed based on the nominal center frequency of the piezoelectric transducer. All local peak points of the current are extracted from the frequency-current response curve. Only peak points whose frequency points fall within the effective bandwidth window are retained. Peak points outside the window are determined to be far-field pseudo-peaks and are removed. The slope morphology recognition includes candidate peak slope calculation, true peak determination, and optimal resonance point locking. For each candidate peak point in the window, the morphology slope of its adjacent frequency points is calculated. Peak points whose morphology slope meets the preset sharpness threshold are determined as effective mechanical resonance points. The frequency point with the largest effective current value among all effective mechanical resonance points is selected as the optimal resonance point of the piezoelectric transducer.

4. The adaptive ultrasonic anti-fouling method based on edge-cloud collaboration according to claim 1, characterized in that: In step S3, the dynamic micro-compensation includes compensation parameter initialization, micro-frequency sweep triggering, resonant point update, and compensation result upload. The edge intelligent terminal performs a small-range micro-frequency sweep within a preset range of the optimal resonant point according to a preset cycle, and collects the effective current value and morphological slope of the micro-frequency sweep point. If there is a frequency point with a larger effective current value and that meets the sharpness threshold, the optimal resonant point is immediately updated to that frequency point, and the updated resonant point parameters are simultaneously uploaded to the cloud-side intelligent management platform.

5. The adaptive ultrasonic anti-fouling method based on edge-cloud collaboration according to claim 1, characterized in that: In step S4, the current time-domain and frequency-domain feature acquisition includes high-speed sampling of the current waveform, time-domain feature extraction, and frequency-domain feature extraction. The high-speed analog-to-digital conversion module continuously acquires the bus current waveform at the optimal resonance point at a preset sampling rate. After filtering the current waveform, the digital signal processing module extracts time-domain features such as the effective value of the current and the peak factor. The time-domain signal is converted into a frequency-domain signal through fast Fourier transform, and frequency-domain features such as total harmonic distortion are extracted. The multidimensional load feature vector is an integrated vector of current RMS value, peak factor, and total harmonic distortion. The end-side intelligent terminal compares the feature vector with the preset load state judgment threshold to make a preliminary judgment on the clean state, initial biofilm state, hard adhesion state, and equipment failure state.

6. The adaptive ultrasonic antifouling method based on edge-cloud collaboration according to claim 1, characterized in that: In step S5, the anti-fouling working mode includes normal anti-fouling mode, early warning enhancement mode, strong stripping mode, and fault shutdown mode. The terminal-side intelligent terminal adjusts the anti-fouling strategy parameters such as output power, drive waveform, and working sequence according to the preliminary judgment result of the load status. The operational data includes device number, optimal resonant frequency, multidimensional load characteristic vector, load status, operating mode, device temperature, measured impedance, and online status. All data is packaged in a standardized manner according to the cloud-side protocol and uploaded periodically via the wireless communication module. When the network is interrupted, it is cached locally and automatically re-uploaded after the network is restored.

7. The adaptive ultrasonic antifouling method based on edge-cloud collaboration according to claim 1, characterized in that: In step S7, the temperature-impedance reference surface is a nonlinear surface model of the piezoelectric transducer constructed in clean water based on temperature, frequency, and impedance. The cloud-side intelligent management platform queries the corresponding theoretical clean impedance in the model based on the equipment temperature and optimal resonant frequency uploaded by the end side, and calculates the residual impedance by the difference between the measured impedance and the theoretical clean impedance, thereby decoupling temperature drift from bioattachment and equipment failure. The cloud-side fusion feature vector is an integrated vector of residual impedance and end-side multidimensional load feature vector. The cloud-side intelligent management platform inputs the fusion feature vector into the machine learning model to achieve secondary determination of load status, and at the same time completes the accurate distinction between individual equipment faults and abnormal bio-attachment.

8. The adaptive ultrasonic anti-fouling method based on edge-cloud collaboration according to claim 1, characterized in that: In step S8, the virtual grid division of the sea area sets the grid scale according to the distribution density of equipment in the sea area, divides the sea area into equal-scale sections according to latitude and longitude, and assigns a unique grid number. The device geolocation mapping is achieved by binding the latitude and longitude coordinates of the end-side device with the corresponding grid number, thus establishing a mapping relationship between the device and the grid. The abnormal data tracing and determination includes the statistics of the number of abnormal devices within the grid, the calculation of the abnormal ratio, and the determination of the abnormal type. Based on the ratio of abnormal devices and the distribution characteristics of faulty devices, it distinguishes between individual device faults, regional biological outbreak risks, and common faults of devices within the grid.

9. The adaptive ultrasonic antifouling method based on edge-cloud collaboration according to claim 1, characterized in that: In steps S1 to S10, the dynamic control of the end-side resonance point and the quantitative decision-making of the cloud-side regional collaborative strategy are realized through the end-cloud collaborative adaptive pollution prevention integrated decision-making formula. The formula expression is as follows: in, This is a comprehensive decision value for edge-cloud collaborative adaptive anti-pollution. As a weighting factor for the effectiveness of end-side resonance, For cloud-side temperature drift decoupling weighting factors, For grid anomaly collaborative weighting factor, The slope of the shape. The slope sharpness threshold. The optimal effective value of the resonant current. For the effective operating current threshold, For residual impedance, The residual impedance threshold, For total harmonic distortion, The harmonic distortion threshold. The proportion of abnormal devices in the grid. The threshold for the grid early warning ratio, This is a correction factor.

10. An adaptive ultrasonic anti-fouling system based on edge-cloud collaboration, used to implement the adaptive ultrasonic anti-fouling method based on edge-cloud collaboration as described in any one of claims 1-9, characterized in that: It includes an edge-side intelligent terminal and a cloud-side intelligent management platform. The edge-side intelligent terminal and the cloud-side intelligent management platform achieve bidirectional data interaction through a wireless communication network. The edge-side intelligent terminal includes a self-test verification module, a frequency sweep acquisition module, a pseudo-peak removal module, a resonance point locking module, a dynamic compensation module, a feature extraction module, a state determination module, a strategy execution module, and a data communication module. The cloud-based intelligent management platform includes a data access module, a temperature drift decoupling module, a secondary status determination module, a grid mapping module, an anomaly tracing module, a strategy generation module, an instruction issuance module, and a feedback optimization module.

Images